Local inference · real measurements

Which GPU for your local LLMs ?

Tokens/s, actual VRAM usage, real power draw: every card is measured with the same llama.cpp protocol. Numbers you can finally compare, before you buy.

3

GPUs measured

21

Verified runs

6

Models tested

Our reference model: Meta Llama 3.1 8B Instruct Q4_K_M — tested by every capable card, it's what makes the leaderboard comparable.

Generation leaderboard

Average tokens/s · select one or more models

GPUVRAMGenerationPromptAvg powerVRAM usedRunsPrice
1RTX 5070 Ti16 Go151 tok/s6 818 tok/s161 W5,4 Go1
2RTX 4090demo24 Go132 tok/s5 900 tok/s285 W6,2 Go1
3RTX 5080demo16 Go124 tok/s5 100 tok/s250 W6,0 Go1
4RTX 4080demo16 Go103 tok/s4 700 tok/s230 W6,1 Go1
5RTX 4070 Ti12 Go89,6 tok/s5 960 tok/s158 W5,6 Go2
6RTX 407012 Go82,3 tok/s4 336 tok/s138 W6,8 Go1
7RTX 3060demo12 Go48,0 tok/s1 500 tok/s155 W5,9 Go1

Generation = how fast the answer is written (the metric you actually feel). A "—" = card not tested on this model (often: it doesn't fit in its VRAM). “Buy” = links to retailers (LDLC, Materiel.net, TopAchat, Cybertek, Grosbill, Rue du Commerce, Amazon).

90 tok/s — what does that actually feel like? Put cards side by side and watch the same answer being written at their real speed.

One protocol, three steps

No manual input: the CLI runs the benchmark and collects telemetry itself. Everything in the database was measured, never declared.

Download

llama.cpp (official binary) and the reference GGUF model.

Run the CLI

One command: it measures tokens/s, VRAM, temperatures and power draw.

Publish

The verified run joins the database and appears in the leaderboard.

Full guide